NVIDIA Nemotron 3 Ultra Released as 550B Parameter MoE Model for Agent Workflows

NVIDIA has released Nemotron 3 Ultra, a 550B-parameter Mixture-of-Experts (MoE) model with 55B active parameters. It is optimized for orchestrating complex, long-running agent workflows by combining frontier reasoning and high throughput with domain adaptability. The model includes architectural innovations like hybrid Mamba-Transformer layers, NVFP4 quantization for cross-architecture GPU deployment, LatentMoE for efficient expert routing, and multi-token prediction for improved generative speed. It is trained using NVIDIA NeMo RL and Gym open libraries with large datasets and is fully open-source, including weights, data, and recipes.